Exclusively from Michael Pettis’ latest newsletter.
Chiwoong Lee at Goldman Sachs has a new report out (“China vs. 1970s Japan”, September 25, 2012) in which he predicts that China’s long-term growth rate will drop to 7.5-8.5%. I disagree very strongly with his forecast, of course, and expect China’s growth rate over the next decade to average less than half that number, but the point of bringing up his report is not to disagree with the details of his analysis.
I want instead to use his report to illustrate what I believe is a much more fundamental problem with these kinds of research pieces on China. The mistake I will argue he is making is one that is fairly common. It involves determining the past relationship between certain inputs and the outputs we want to forecast – say GDP growth. Once these are determined, the economist will carefully study the expected changes in the inputs, and then calculate the expected changes in the outputs, to arrive at his growth forecast.
This is pretty much the standard analysis provided by the IMF, the World Bank, and both academic and sell-side research, but, as I will argue, this methodology implicitly assumes no real change in the underlying development model – no phase shift, to use a more fashionable term. If this assumption is correct, then the analysis is useful. If however we are on the verge of a shift in the development model – perhaps, and usually, because the existing model is unsustainable and must be reversed, the analysis has no value at all.
Lee arrives at his 7.5-8.5% range by comparing China today with Japan in the early 1970s. He considers reasonable and very plausible changes in various inputs and concluding on that basis that Chinese growth will slow from the torrid levels of the past decade, but will nonetheless exceed the roughly 5% real growth rate achieved by Japan in the two decades following the early 1970s. I assume his Chinese growth prediction is also for the next one or two decades but I was not able to determine if this is in fact the case.
What about China? Like Japan, (1) real wages and the labor share of income are rising in some areas and (2) the pace of technological advances has surpassed its peak. However, (3) while consumer durables are spreading, there is still ample room for growth, (4) the export ratio is high due to economies of scale, and (5) growth remains high despite an increase in raw material prices reminiscent of the oil crisis.
Although both potential and actual growth is expected to remain high in China, gradual decline is likely from the double-digit (%) pace of 2004-2007. Our China economics team calculates a range of 7.5~8.5% for China’s potential growth.
As the excerpt above suggests, Lee focuses on six input factors specifically: the shift from labor glut to labor shortage, the pace of technological catch-up, the spread of consumer durables, economies of scale and export structure, the impact of raw material price rises, and political stability. He compares the impact of changes in each of these on the Japanese economy as a source of the Japanese slowdown, and then estimates their comparable impact on the Chinese economy. This allows him to estimate the amount of the expected slowdown in China.
The piece is a very interesting one, and it is well worth reading for the information and insights it provides, but in my opinion it shares a fundamental problem with nearly all of the other analyses that compare China today with Japan in the 1960-70s, rather than Japan in the late 1980s. Many of these analyses provide much less sophisticated reasons than Lee’s. For example the most popular reason for comparing China with Japan of the 1960-70s is that China today is much poorer than Japan in the late 1980s. Japan in the late 1980s was rich, people will say, while China is terribly poor, so there can’t be any useful comparison between Japan in the 1990s and China in the next decade.
What are the real similarities?
This, of course, is silly – and not just because it implies only rich countries can have crises. If you are arguing about the consequences of imbalanced, investment-driven growth, it isn’t the nominal levels of wealth that need to be compared. After all there are rich as well as poor countries that suffered from this kind of unbalanced, investment-driven growth, and all of them ended up suffering subsequently from the same kinds of economic rebalancing.
What really matters is the extent of the underlying imbalances and the relationship between capital stock and worker productivity. In that light it is just as easy for a poor country to have excess capital stock as it is for a rich country – perhaps even more so.
Lee of course doesn’t fall into this trap, but he does make the same mistake, I think, that all these other analyses make. He focuses on nominal variables in each of the two countries and compares them, trying to find similarities not in the level of imbalance but rather in income and development levels. He provides, for example, this very interesting table:
China today, as the table shows, shares many development characteristics with Japan in the late 1960s and early 1970s, although it is still poorer than Japan was.
But in this case it is clearly not the variables that measure the level of development that matter. We should instead be focusing on the extent of domestic imbalances.
Why? Because focusing on comparable development levels is useful only if we assume that China’s future is going to look a lot like China’s recent past. We are implicitly assuming the development model in China will remain largely unchanged. In this case, of course, a little more or a little less of any particular input should have a more-or-less predictable impact on subsequent growth rates.
Where this approach is useless is when the growth model generates domestic imbalances that become increasingly unsustainable, in which case any long-term forecast must assume that the existing growth model will be abandoned and replaced by another growth model, one that allows the underlying imbalances to reverse themselves.
Because this process is path dependent, and usually subject to important political constraints, it is hard to predict exactly when the old growth model will be replaced by a new growth model (for example I did not believe that China’s rebalancing would begin until 2013, after the new leadership took power, but it may actually have begun in 2012). It is also hard to predict short-term consequences, although it is, perhaps paradoxically, much easier to predict the medium and long-term implications.
Why? Because it is almost axiomatic that unsustainable imbalances must reverse themselves one way or another, and the only interesting question is how. The reversal of major imbalances is almost always very difficult, but the process itself can occur either in a quick and “catastrophic” way, via a kind of sudden financing stop that may lead to a financial crisis and negative growth, or in a slow, more controlled grinding away of the imbalances. There are few other ways in which the rebalancing can occur once the imbalances have gone far enough.
This is why I think it is much more appropriate to compare China today with Japan in the late 1980s. Japan in the late 1960s and early 1970s may share many developmental characteristics with China today, but the Japanese economy in that earlier period never achieved, as far as I can see, the kinds of imbalances that it did much later in the 1980s, and so it is not really comparable to an extremely unbalanced China today.
These imbalances are well documented. The important characteristics of Japan in the late 1980s would almost certainly include the following:
- Japan in the late 1980s grew at extraordinary rates fueled by a credit-backed investment boom funded at artificially low interest rates.
- Although for many decades much of the investment may have been viable and necessary, by the 1980s investment was increasingly misallocated into expanding unnecessary manufacturing capacity, as well as fueling surges in real estate development and excess spending on infrastructure.
- Artificially low rates, set nominally by the central bank but in reality by the Ministry of Finance, and coming mainly at the expense of household savers also fueled a bubble in local assets.
- An artificially low currency fueled very rapid growth in the tradable goods sector while also constraining household income growth.
- Because the growth model constrained growth in household income and household consumption, it forced up the domestic savings rate to extraordinary levels.
- The combination of low consumption and excessive manufacturing capacity required a high trade surplus in order to balance production with demand.
- And finally, and most worryingly, debt levels across the economy began to soar as debt rose much faster than debt servicing capacity.
All of this is true of China today, and this is why it is much more important to understand how Japan rebalanced after 1990 if you want to understand the challenges and risks facing China today. China is not like Japan in the 1950s, 1960s or 1970s in any meaningful way even if its current development level is much closer to Japan during those decades. Because of the serious imbalances China is much more like Japan in the late 1980s, with the major difference being that Japan never took debt, investment, and consumption imbalances to anywhere near the levels that China has taken them.
For this reason what we really have to consider when thinking about China is how these imbalances tend to be reversed. Since they were reversed in Japan in the period following 1990, Japan provides at least one possible model for China’s rebalancing process and, perhaps much more importantly, it demonstrates the kinds of pressures that China will face as it is forced into rebalancing.
This insight, by the way, extends to a lot more than just Chinese economic growth over the next decade. It is applicable whenever a system that has generated unsustainable imbalances is in the process – as it eventually must – of reversing these imbalances. In that case it is a waste of time to extrapolate from the development of variables during the period in which the imbalances were created to the period in which the imbalances are reversed.
With a change in the growth model comes a radical change in the relationship between underlying variables and their impacts on growth. It makes no sense to use the earlier data series, adjusting them according to new conditions, and to project new data series, because when a country is forced into reversing the imbalances, by definition all the correlations between relative inputs and outputs must fall apart, and the more extreme the imbalances that need to be reversed, the more untrustworthy the previous relationships. In that case it is much more useful to posit the various ways in which a reversal of the imbalances must occur.
Economic growth in Europe over the next ten years, for example, will not be anything like economic growth in the last ten years adjusted for changes in demographics, taxes, or anything else. We will be dealing with a Europe in which the tremendous debt, currency, and labor cost imbalances of the past decade must be reversed, and since the most likely form of the reversal will entail the breaking up of the euro, any growth predictions that do not at least acknowledge this chaotic rebalancing process are likely to be flimsy at best.
Likewise with predictions of US consumption growth, which will have to deal with reversals in the savings and consumer credit trends of the past decade, and with sharp change in two decades of housing behavior. Predictions about the prices of hard commodities, which have to consider a major dislocation in the source of commodity demand since the early part of the last decade, are also likely to be highly unstable.
If you want to make economic predictions, in other words, whereas a long historical view will be very useful because it allows you to consider the dislocations created by a reversal of unsustainable imbalances, recent economic data are largely useless, as are predictions based on linear adjustments of recent economic data. Instead of projecting from past data you must model the various paths by which rebalancing can occur, and your prediction must be limited to those paths.